The primary aim of this meta-analysis was to produce a comprehensive modelling of the dose–response relationships between resistance exercise and commonly used measures of physical performance within S&C. The analyses identified that a range of factors are associated with the magnitude of change across resistance-only and resistance-dominant training interventions. These factors include the length of the intervention, outcome type, volume, overall intensity of training, and the intensity of load. The analyses also identified important interactions between loading intensity and outcome domain, such that some outcomes are more likely to experience greater improvements with a range of sub-maximum loads (30–70% 1RM).

Length of intervention has generally not been explored in dose–response modelling due to most reviews focusing on a smaller number of homogeneous studies with a restricted range of durations. Across the studies included in this meta-analysis, durations were found to be relatively short, with the median duration equal to 8 weeks and 97% of interventions lasting less than 26 weeks. The prevalence of shorter duration studies may be due to multiple challenges associated with longer duration interventions including an increased need for resources as well as greater inclusion and agreement from key stakeholders (athlete, organisation and management), increased dropouts, and scheduling difficulties to fit within non-competitive periods [29]. Despite the relatively short intervention durations, the results show that substantive improvements can be made, and that longer durations within these time frames create greater mean improvements. Therefore, practitioners and athletes could use shorter time periods, potentially during off-season or preseason, to maximise physiological improvements when competition or sporting demands are lowest. The effect of duration remained consistent throughout the model-building process with standardised mean differences estimated to increase by approximately 0.03 for each additional week of training. Given the relatively short and homogeneous durations included, there was limited ability to explore the functional form of changes over longer durations (e.g., playing seasons, years). In a recent large modelling study investigating the time-course of strength adaptations, Steele et al. [30] showed that linear-log growth models were appropriate to describe improvements of relatively untrained participants over the course of almost 7 years, with improvements tending to plateau after approximately 1 year. The training stimulus investigated by Steele et al. [30] focused on minimal dose resistance training (1x/week, single sets to momentary failure of six exercises), which is likely to have influenced the parameters obtained. Our analysis was limited to durations of no more than 26 weeks and thus we cannot draw inferences as to how results might change over longer time frames. Further research is required to better understand the influence of duration of an intervention and likely interactions between participant characteristics, the specific outcome, the training stimulus, and changes in the training stimulus following, for example, different periodized approaches.

The manipulation of acute program variables within resistance training interventions is often focused on the development of a single outcome domain. Previous meta-analyses investigating dose–response relationships have predominantly focused on the development of muscular strength and/or hypertrophy [7,8,9,10,11,12,13,14,15,16,17]. The current meta-analysis demonstrated varying effects across multiple outcome domains commonly targeted with the greatest effect sizes obtained for strength, and the lowest obtained for sprint performance. Resistance training for the purpose of improving maximum strength is arguably the most investigated and well-understood area within S&C [1, 2], and greater effect sizes may reflect this increased refinement and specificity between traditional resistance-training methods and maximum strength outcomes [1]. In addition, researchers frequently test maximum strength using the same exercises included in the training intervention, further increasing specificity and potentially the improvements measured [31]. Results from a previous meta-analysis indicate that the dose–response relationship between the % 1RM and strength gains diminishes when testing is carried out isometrically [15]. Practitioners should be advised to select an appropriate testing mode for the given training stimulus applied, whilst being aware of any potential upward or downward shift in expected results dependent on the measurement used. Further study is needed to provide greater context to the transfer of strength from varied magnitudes of load to neutral testing modalities.

Following maximum strength, jump performance and power generated the next largest effects. Similar magnitude improvements in jump performance and power may be expected given the well-established relationship between the two factors [32,33,34,35]. In addition, many of the studies measured power during loaded and unloaded jumps, further increasing associations and similar magnitude improvements. Outcomes relating to sprint performance demonstrated the lowest magnitude improvements. Sprinting comprises a substantial and complex technique element [36,37,38,39,40], and given the relatively low number of studies (~ 12%) that included sprint specific interventions, a lower effect size distribution may be expected. More broadly, the lower effect size distribution for sprint outcomes may also reflect a lack of specificity with regards to development of relevant physical outputs. Transference between improvements and long-term adaptations in S&C is dependent primarily on the training principles of specificity and progressive overload, respectively [41, 42]. Most training methods included in the meta-analysis focused on bilateral production of maximum vertical forces over long durations. In contrast, sprinting activities require high forces produced over short ground contact times that are predominantly unilateral with substantive horizontal components in relation to the body’s position relative to the ground [37, 38, 43, 44]. In a recent meta-analysis, Murphy et al. [4] showed moderate to strong relationships between improvements in strength, power and sprint performance in team sport athletes, concluding that greater development of physical capacities may result in further improvements in sprint performance. Despite these correlations, however, researchers have also shown that large increases in maximum strength (~ 12–18%) translate into only small decreases in sprint times (~ -2–8%) [45,46,47]. With evidence to suggest restrictions may exist in the transference of physical capacities to sprinting, further sprint-specific training modes such as resisted sprint training may provide additional benefits allowing for the ability to overload kinetic output with increased kinematic specificity to the complex movement of sprinting [48]. Collectively, there appears to be scope for future research to investigate why improvements in sprint performance are generally much smaller than other outcome domains and whether this difference can be ameliorated with a focus on certain training practices.

Movements associated with COD and agility could be considered even more complex than sprinting due to the high acceleration and deceleration demands, the ability to rapidly alter body position, combined with the need in some activities to react to an external stimulus [49, 50]. The results of the current meta-analysis suggest improvements in COD are likely to be of similar magnitude to those measured during vertical jump and tasks focusing on the development of power, albeit with a greater level of uncertainty. Outcomes measuring COD and the related construct of change of direction speed and agility represent a developing area within S&C [51, 52] with only ~ 3% of outcomes assessing COD performance. Whilst reasons for the larger effect size distribution in comparison to linear sprinting require further study, potential explanations include the complex, multifaceted nature of the tasks and the scope for multiple limiting factors to be addressed. Additionally, it is recognised that many agility and COD tasks include substantive skill elements [52], such that failure to appropriately familiarize participants could lead to systematic biases in regard to learning effects and subsequent overestimations of effect sizes.

The results of the current meta-analysis demonstrate the importance of training intensity. Previous researchers investigating dose–response relationships have tended to contextualise and quantify intensity based on load and thereby % 1RM [7, 8, 14, 15]. This approach works best when considering traditional strength or hypertrophy focused interventions comprising large compound movements where 1RMs can be measured and appropriately summarise a relevant feature of intensity [53]. An aim of the current meta-analysis was to investigate dose–response relationships across a range of resistance-based training modalities and outcomes; therefore, in addition to intensity of load expressed as a % 1RM, a more general categorisation scheme was included. Interventions comprising predominantly ballistic, loaded jumping or sprinting exercises were always considered high intensity, due to the high mechanical loads and assumption that they are conducted with maximal intent, unless stated otherwise. Across all outcomes, evidence was obtained that greater overall intensity was associated with increased effect sizes, with interventions judged to be of medium and high overall intensity expected to increase effect sizes by approximately 0.10 relative to low overall intensity. When prescribing intensity, practitioners should consider training intensity beyond simply intensity of load (% 1RM). For example, ensuring maximum intent during ballistic exercise exemplifies a simplistic method of prescribing higher intensity training. Incorporating methods such as velocity-based training provides the opportunity for instantaneous feedback during ballistic movements to encourage maximal intent [54]. Adjusting the intensity of plyometric training may be more complex and largely dependent on exercise selection to increase or decrease take-off and landing ground reaction forces [55].

Additional detailed information on the dose–response relationship of intensity was obtained when investigating potential interaction effects between outcome and intensity of load measured by % 1RM. The results identified a range of different profiles, with no clear pattern for COD, monotonic increases for strength and speed, a monotonic decrease for jump performance, and a parabolic profile for power. The best estimate profile for maximum strength appeared non-linear with an inflection point ~ 70% 1RM, where further increases in effect size estimates started to slow with additional load. The results of the present meta-analysis are consistent with previous reviews, e.g. Peterson et al. [8] identified increased effects with heavier % 1RM loads but diminishing effects, particularly with untrained participants [18]. The results also align with previous research indicating that heavy load training may increase muscle activation by up to 30% [56], conceivably providing a stronger stimulus for adaptation. Some authors have also suggested, however, that improvements in strength with greater loads may be inflated due to high specificity of task and outcome, given that strength testing is often conducted performing the 1RM of the movement being trained [15].

Analysis of sprint performance also identified a monotonic increase in effect, with the greatest increases obtained with the heaviest loads. The most common sprint outcomes investigated in S&C research include the time to sprint between 5 and 50 m, with the most frequent intervals comprising 10, 20 and 30 m [4]. Most studies have been conducted with either team sport or untrained participants who achieve maximum velocity between 15 and 40 m, in comparison to trained sprinters who require distances of 40–80 m to achieve maximum velocity [38, 44, 57, 58]. Consequently, sprint data collected over 10–30 m may provide researchers with divergent outcomes describing both acceleration and maximum velocity. Previous studies have reported strong associations with outcomes designed to assess acceleration (e.g., 10 m) and horizontal force, power and relative strength with longer duration ground contact times (approximately 200 ms) [37, 38]. In contrast, maximum velocity sprinting has been shown to be dependent on the ability to maintain large horizontal and vertical forces whilst minimizing braking forces with reduced ground contact time (approximately 100 ms) [38, 39, 58]. Previous meta-analyses have concluded that high-intensity non-specific resistance exercise is among the most effective training methods to improve sprint performance in team sport athletes [4, 36, 59]. Neural and morphological adaptations associated with high-load resistance exercise and improved force output may provide a mechanism for positive transfer to the high levels of horizontal force required during early-phase acceleration to improve sprint performance. Highly trained individuals may require more specific training methods, however, that target improvements in physical qualities while matching the kinematic demands of sprinting [1, 4].

In contrast to the increasing dose–response relationships with intensity for strength and sprint performance, results identified monotonic decreases for jump performance with the largest effects obtained at ~ 30% 1RM. Jump performance is dependent on net impulse and take-off velocity such that % 1RM loads lifted with maximum intent provide sufficient stimulus but do not limit velocity to a large extent may provide the greatest transfer to improvements in jump performance [33, 60]. Previous researchers have also demonstrated that jump squat training with low (< 30% 1RM) or no additional load can produce velocity-specific adaptations associated with improvements in jump performance [61]. The intensity profile for outcomes measuring power production was parabolic, with the greatest improvements obtained between ~ 40 and 70% 1RM. These results support the hypothesis that performing resistance exercise with loads that elicit the largest power outputs is the most effective method to improve power and that for most exercises power is maximised between 30 and 70% 1RM [62]. During weightlifting exercises (clean, snatch, hang and pull variations), power is maximised with heavier loads (≥ 70% 1RM), whereas loads of 0–30% 1RM maximise power during jump squat exercises [62]. As the optimum load for power production is exercise-dependent, practitioners should be aware of the appropriate load required to stimulate peak power output within the prescribed exercises and endeavour to create athlete-specific profiles where access to relevant measurement devices is available.

The influence of training frequency and volume on strength and hypertrophy has been assessed in several previous meta-analyses [10,11,12, 16]. The results obtained herein were mixed but showed limited evidence that these factors were influential. No improvement in model performance was obtained when including the number of sessions per week as a categorical or continuous predictor, or the average number of exercises per session. Only limited evidence of model improvement was obtained for the average number of repetitions per set, with the marginal effect showing declines as the number of repetitions increased. In contrast, evidence was obtained for greater effect sizes with increasing number of sets per session. Seminal research by Rhea et al. [9] was among the first, within S&C, to use meta-analytical techniques to assess the use of single versus multiple sets in resistance exercise for strength development. The authors concluded multiple sets were more beneficial than single-set training. A follow-up meta-regression by Krieger et al. [63] found a 46% increase in muscular strength when completing two to three sets, in comparison to single sets, although no further difference was found for resistance exercise with more than four sets. More recently, researchers have concluded that increasing weekly training volume through increased number of sets can produce similar results to increasing training frequency [11, 12]. The current meta-analysis demonstrates that focussing on a smaller number of key exercises while completing multiple sets at an appropriate intensity for a targeted outcome may be more beneficial than attempting to perform many exercises with an increased frequency.

The training status of participants is a key consideration when designing and implementing resistance exercise. Previous meta-analyses have demonstrated rank-order effects, with the largest improvements obtained by untrained participants [1, 7], followed by recreationally trained and then highly trained participants. In contrast, the current meta-analysis found a lack of evidence to support different effects across the training status categories. Differences in results obtained in the present versus previous meta-analyses may be due to several reasons. Previous analyses have been less formal than those conducted herein, with authors identifying differences based primarily on point estimates. In contrast, participant training status was assessed in the present analysis with predictor variables for lower levels already included in the model and addition of the factor was assessed based on ability to improve model performance. The lack of data for highly trained participants, disproportionate inclusion of untrained participants, combined with short-duration (≤ 26 weeks) interventions, which are known limitations within S&C research [64], may have also influenced the results obtained and discordance with what is generally believed in the field. With advancements in technology and ability to collect valid and reliable high-frequency data over longer periods across all levels of sport and recreation, use of longitudinal data collected in the field provides opportunities to better investigate differences in dose-responses relative to participant training status.

In addition to training status, the current meta-analysis found a lack of evidence to support different effects between sexes. Although males exhibit greater levels of baseline strength and muscle mass [65], the current meta-analysis results are consistent with previous research that has failed to identify any difference in effects between sexes in improvements in strength or hypertrophy, with indications that varying levels of adaptations may be more related to relative strength [